This work deals with the clustering of information sources for epipole estimation in a multi-camera system. For this problem, each pair of matched visual features in the images can be considered as an elementary information source. The epipole is then estimated by combining these elementary sources taking into account their inadequacy, in particular large imprecision and presence of outliers, as well as the very large number of sources. We address the challenges introduced by a large number of sources with a strategy based on clustering and intra-cluster fusion using the Belief Functions framework. When evaluated on real data, the proposed algorithm exhibits more robustness in terms of accuracy and precision than the standard approaches which provide singular solutions.

这项工作涉及多相机系统中对极估计的信息源聚类。对于这个问题，图像中每一对匹配的视觉特征都可以被认为是一个基本的信息源。然后通过结合这些基本来源，考虑到它们的不足，特别是大的不精确性和异常值的存在，以及非常大量的来源，来估计对极。我们通过使用信念函数框架的基于聚类和聚类内融合的策略来解决大量来源带来的挑战。当对真实数据进行评估时，所提出的算法在准确性和精度方面表现出比提供奇异解的标准方法更强的鲁棒性。